Improve readability and efficiency for ZUNION operation

[QuChen88]

Currently zunion operation in t_zset.c is using dictAddRaw() to search for the element in the accumulating dictionary. If the element was not found and got added, it would then set the key to a new key string via dictSetKey(), which is counter intuitive and not very effective.

This PR improves the code readability and effectiveness by using dictFind() to search for the element in the dictionary, and then uses dictAddRaw() and dictSetDoubleVal() to create the element in the set, and avoids the need to call dictSetKey().

Note - this change is consistent with the original implementation of sorted set

valkey/src/t_zset.c

Line 2286 in d680eb6

de = dictFind(accumulator,zuiSdsFromValue(&zval));

. It was later changed to a "not-so-nice" style in commit 68bf45f

[ranshid]

Thank you @QuChen88 I think the zset code is about to change a lot following #1427. maybe you want to take a look and propose alternatives there?

[JimB123]

I think the reason for the original code is that in a single operation, it searches and adds. With the new code, it will need to perform a hash table lookup for the find, and then another hash table lookup for the add.

[QuChen88]

My biggest issue was that the insertion into the dict happened outside of the if (existing) check previously.

The new code in the proposed PR #1427 looks correct to me. It first finds the element in the hashset and then if it is not present will add the element.

[zuiderkwast]

The new code in the proposed PR #1427 looks correct to me. It first finds the element in the hashset and then if it is not present will add the element.

Great, so let's close this PR? #1427 will be merged very soon.